#
✕ nick:sven
#asahi-dev · 2022-03-20 11:54
svenyeah, something like full desktop environment + compiling the kernel + find /
#asahi-dev · 2022-03-20 11:52
sventhat would be great! can you first try to just revert https://github.com/AsahiLinux/linux/commit/af81cf607423e842a6a0146e8cd35cbaf6aab178 on the asahi branch and see if nvme still works?
#asahi-dev · 2022-03-20 11:41
sveni'm glad they don't get shipped overnight. otherwise i'd have my excuse now :D
#asahi-dev · 2022-03-20 11:41
svenso far I've managed to resist buying that studio ;)
#asahi-dev · 2022-03-20 11:39
svenmarcan: ^--
#asahi-dev · 2022-03-20 11:38
svendon't think so
#asahi-dev · 2022-03-20 11:37
svenbut then again, I can't reproduce this on my M1 at all. maybe it's just something strange that only happens on the Max/Pro
#asahi-dev · 2022-03-20 11:36
sven(race condition as in "you can't submit a command when you do X with the nvme controller at the same time")
#asahi-dev · 2022-03-20 11:35
svenbut I wasn't able to find any race
#asahi-dev · 2022-03-20 11:35
svenmy other guess was that it was some kind of race condition and the lock just made it unlikely to lose that race
#asahi-dev · 2022-03-20 11:33
sveniirc the bug was then that the command appeard on the completion queue but no interrupt was triggered. it was only picked up after the 30 second timeout by polling that queue.
#asahi-dev · 2022-03-20 11:32
svenhrm, not sure there are any. I just submit a command there and then just wait until the controller puts it to the completion queue and fires an interrupt. i don't even care if the controller only sees the write seconds later.
#asahi-dev · 2022-03-20 11:27
svenohh.. okay. let me think what the consequences would be if the writel happens far down the line
#asahi-dev · 2022-03-20 11:22
svenaxboe had a system that was affected though, maybe he can give it a try
#asahi-dev · 2022-03-20 11:22
sveni'm also not sure I completely understand. what do I want to order against with an additional dmb after the writel?
#asahi-dev · 2022-03-20 11:21
sveni can't reproduce it on my systems unfortunately which makes this so much harder :(
#asahi-dev · 2022-03-20 11:17
sventhat writel tells it "please now start executing the command at index i in this array"
#asahi-dev · 2022-03-20 11:17
svenit can execute other commands
#asahi-dev · 2022-03-20 11:15
svenand without the lock on some machines the command still executes just fine (so the nvme controller saw it) but never generates a completion interrupt
#asahi-dev · 2022-03-20 11:14
svenso the weird thing that all I should need at that point is the guarantee that the memcpy before is visible to the nvme controller. that writel there writes to a hardware FIFO anyway that should be large enough
#asahi-dev · 2022-03-20 10:32
sveni still don't understand why that lock is required. and i can't even reproduce the errors without that lock on my mini :/
#asahi-dev · 2022-03-20 10:31
svenhrm, and i guess that command issue lock could use a description
#asahi-dev · 2022-03-20 10:30
svenj`ey: let's see if the bot is happier with this https://github.com/AsahiLinux/linux/commits/nvme-rfc. just need to finish the cover letter now
#asahi-dev · 2022-03-20 08:56
svenand I think bitfield.h instead of bits.h this time
#asahi-dev · 2022-03-20 08:56
svenyeah, and restricting rtkit to 64Bit only
#asahi-dev · 2022-03-20 08:54
svenand then maybe wait another ~5 hours or so to give it another chance to scream at me ;)
#asahi-dev · 2022-03-20 08:53
sventhe kernel bot started screaming at me after I fixed that compile test makefile thing so I’ll have to check what I need to fix there
#asahi-dev · 2022-03-20 08:53
svenmaybe
#asahi-dev · 2022-03-19 15:17
svenprobably. my joke was a reference to the academic ones i had to suffer through though! :)
#asahi-dev · 2022-03-19 15:16
sventhanks, fixed!
#asahi-dev · 2022-03-19 15:16
sventhat's obviously just a spelling mistake for reviewer #2 to catch ;p
#asahi-dev · 2022-03-19 15:03
svenj`ey: ^--
#asahi-dev · 2022-03-19 15:03
sven~95% done with what i'll submit for nvme, https://github.com/AsahiLinux/linux/commits/nvme-rfc
#asahi-dev · 2022-03-19 14:46
sventhe one thing i'm still unhappy about is this mystery lock required around command submission (but not for me!): https://github.com/AsahiLinux/linux/blob/asahi/drivers/nvme/host/apple.c#L329
#asahi-dev · 2022-03-19 14:45
svenalright, "apple,tXXXX-nvme-ans2", "apple,nvme-ans2" it is then
#asahi-dev · 2022-03-19 14:42
sveni assume we're all happy with "apple,tXXXX-nvme-ans2", "apple,nvme-ans2"?
#asahi-dev · 2022-03-19 14:13
svenkettenis: does openbsd or u-boot care about the sart compatible already?
#asahi-dev · 2022-03-19 14:12
svenmeh, i guess it'll be yak shaved on the ML anyway ;)
#asahi-dev · 2022-03-19 14:12
sven*sart3 actually
#asahi-dev · 2022-03-19 14:11
svenso "t8103-sart" but "t6000-sart", "sart2"?
#asahi-dev · 2022-03-19 14:08
svenhuh, looks like SARTv2 might actually be a A14 / M1 exclusive thing
#asahi-dev · 2022-03-19 14:00
svenalright, SART1 was 36 bit address space and filter entry length 24 bit then. SART2 just increased the filter entry size to 36 bit. SART3 pushed both to 42 bit.
#asahi-dev · 2022-03-19 13:59
svenhow much address space does the pro/max support again? 42bit?
#asahi-dev · 2022-03-19 13:51
svenoh, also, apple sometimes calls SART coastguard for some reason
#asahi-dev · 2022-03-19 13:50
svenactually, let me figure out how long sart2 was used
#asahi-dev · 2022-03-19 13:50
svenyeah
#asahi-dev · 2022-03-19 13:49
sven1 -> 2 -> 3 is always about adding more paddr space
#asahi-dev · 2022-03-19 13:49
svensart1 is even older
#asahi-dev · 2022-03-19 13:49
svenyeah
#asahi-dev · 2022-03-19 13:49
svenalright, SART bindings: do we want the apple,sart2 and apple,sart3 generics there?

plus ›